杏吧原创

WHEN scientists reported they had created a space-time wormhole in November last year, the world鈥檚 media were all over the story, even though they struggled to make sense of it. A journalist for the website UNILAD put it neatly when they wrote: 鈥淪o, you might have to bear with us here a bit, because it鈥檚 all very complicated and new.鈥�

As far as many observers could see, physicist at the California Institute of Technology and her colleagues had in fact merely used a quantum computer to simulate a wormhole. Good luck flying a spaceship through that. What confused matters was that the team insisted the work amounted to more than just a simulation. The quantum computation, the researchers said, was fully equivalent to the creation of a wormhole.

If you find that hard to swallow, you aren鈥檛 alone. Ask other physicists about Spiropulu鈥檚 claims and you tend to get a lot of long pauses, chin-stroking and disagreement. It seems there is genuine confusion about if and when a quantum computation can create real entities or just simulate them.

The putative wormhole isn鈥檛 the only thing said to have been conjured up by quantum computers recently 鈥� there is also the alluringly named time crystal, as well as strange particles called nonabelions, touted as the ideal ingredient for next-generation quantum computers. But whether these amount to instances of true creation or not is a question that takes us into deep waters. It is a new twist on the riddle that has haunted physics since quantum mechanics was devised in the early 20th century: what is truly real?

Regular computers use transistors as switches to encode the bits, or 0s and 1s, of binary code. We can use them to simulate all kinds of objects and processes 鈥� but it would be absurd to suggest that a simulation of a molecule or a weather system, say, really creates those things. So why should quantum computers be any different?

Quantum computers

In place of encoded 1s and 0s, quantum computers instead use quantum bits (or qubits). These are typically quantum particles, like photons or cold atoms, whose properties are described in the language of quantum mechanics, using an abstract mathematical object called a wave function. Quantum rules permit qubits to be entangled with one another, so that what happens to one apparently influences what happens to others. That is generally how this type of computation can speed up certain kinds of calculation. In other words, the qubits inside a quantum computer are described by the same theory that explains the fundamental particles that make up everything. This is what motivated the proposal for such computing first made by Richard Feynman in 1981. Why not, he said, simulate physical systems using the same quantum rules that govern those systems themselves, instead of clumsy approximations with conventional bits?

To see why this blurs the boundaries between simulation and reality, let鈥檚 put wormholes aside for a moment and look at another recent experiment using quantum computers. It concerns a particle called a non-Abelian anyon (or nonabelion for short). These are a variety of anyon, a kind of particle first hypothesised by theoretical physicist Frank Wilczek in 1982. Anyons have odd properties in between those of the two normal classes of fundamental particle: bosons, which carry forces, and fermions, which constitute matter.

In 1997, physicist Alexei Kitaev showed that hypothetical nonabelions can be moved around each other (or 鈥渂raided鈥�, as physicists put it) to preserve a memory of those movements in a way that could encode quantum information robustly. This means nonabelions could act as qubits that aren鈥檛 prone to the kind of random errors that bedevil existing quantum computing. Kitaev argued that such nonabelions could be created as 鈥�quasiparticles鈥� in certain materials 鈥� that is, as collective emergent states of the electrons in these materials.

Rethinking reality: Is the entire universe a single quantum object?

In the face of new evidence, physicists are starting to view the cosmos not as made up of disparate layers, but as a quantum whole linked by entanglement

Despite several claims to the contrary, there is no compelling evidence that such nonabelion quasiparticles have been created in real materials. But in June, teams of researchers at Google Quantum AI in Mountain View, California, Zhejiang University in Hangzhou, China, and quantum computing company Quantinuum鈥檚 labs in Germany and Colorado claimed to have in quantum computers. They coaxed the qubits into states that have wave functions corresponding to the strange predicted properties of nonabelions. But was this an act of creation 鈥� or just simulation?

Imagine, say, that we want to model a hydrogen molecule on a quantum computer. The molecule is made of two protons bound by a cloud of electrons. A quantum simulation may have exactly the same wave function as these components, but it clearly isn鈥檛 the same as the real-world object. We know what a hydrogen molecule is, and it鈥檚 not a collection of qubits.

This is the framing used by at Google Quantum AI to explain why he thinks the team really 鈥渕ade鈥� nonabelions in a quantum computer chip. When you ask if a simulation is 鈥渞eal鈥�, he says, the question is whether what the quantum computer is doing has a real-world equivalent to be 鈥渕apped onto鈥�. If so, then you have a simulation. But if you are studying a more abstract phenomenon, say, quantum entanglement, then generating it using qubits doesn鈥檛 lack anything present in that phenomenon itself. 鈥淓ntangling two qubits in a quantum processor does not involve mapping their states to that of any system distinct from the processor itself,鈥� he says.

Nonabelions

Nonabelions, meanwhile, are hypothetical particles that 鈥� as far as we know 鈥� don鈥檛 exist in reality at large. So, for Zalcman, there is nothing to be mapped onto. The only way we can define a nonabelion is based on its quantum properties. 鈥淭he process that occurred on the chip is non-Abelian braiding and we therefore say that our investigation is a realisation, not a simulation,鈥� says Zalcman.

Computer scientist at the University of Texas at Austin also alights on this issue of mapping. 鈥淚t depends on what sort of 鈥榯hing鈥� we鈥檙e talking about,鈥� he says. A computer simulation of a hurricane 鈥渄oesn鈥檛 make anyone wet, but a computer simulation of multiplication is multiplication鈥�, he says.

Wilczek also conceived of an entirely different kind of quantum object that now features in this debate. In 2012, he was looking for a way to spice up a university course he was teaching on the structure of crystals, whose atomic arrangement repeats periodically in space. He speculated about a crystal that repeats not in a spatial dimension, but in time. The components of these time crystals would change with the proverbial ticking of the clock (moving around, say), but return to their original state at regular intervals 鈥� forever and ever.

Although Wilczek鈥檚 original notion of time crystals proved impossible to realise, a subset of them, called discrete time crystals, did seem to be feasible. In 2021, teams of scientists at Stanford University in California, Google Quantum AI, the Max Planck Institute for Physics of Complex Systems in Dresden, Germany, and the University of Oxford using Google鈥檚 Sycamore quantum-computing processor. The states that were produced in the qubits displayed exactly the periodic behaviour in time predicted for quantum discrete time crystals.

Time crystals

So was it a real time crystal? 鈥淲e spent a while grappling with this question,鈥� says , a member of the Stanford team. Quantum computing 鈥渋s blurring these lines鈥�, he says, 鈥渁nd there is ongoing debate in the community about terminology.鈥� But he adds that there is a distinction between simulating a quantum time crystal on a classical computer and what his team did. 鈥淚n the former case, a physical system 鈥� the classical computer 鈥� evolves in a way that looks nothing like the physics we鈥檙e interested in, but nevertheless outputs numbers that replicate the outcomes of a hypothetical experiment.鈥� Yet using a quantum computer in the way the researchers did 鈥渋s itself an experiment鈥�, he says, because it involves quantum objects (the qubits) doing what they are supposed to in a time crystal.

Quantum information theorist at the Massachusetts Institute of Technology thinks similarly. The question is whether the qubits are at all times doing what is expected of the system they are modelling. In the nonabelion case, they do, he says. But if they don鈥檛, it is a simulation.

Physicist at the University of California, Berkeley, a specialist in collective quantum behaviour, says he finds the distinction between quantum simulation and quantum reality tough. 鈥淚t鈥檚 all shades of grey,鈥� he says. For him, it comes down to a question of usefulness. The reason we are interested in room-temperature superconductors, say, is to transport electricity without losses or to levitate high-speed trains. But if you got a bunch of qubits into the wave function corresponding to a superconducting state, you could never use it to do any of those things. The nonabelions, on the other hand, 鈥渁re probably just as useable鈥� for error-resistant, or topological, quantum computation as making them in some solid-state material, says Yao. 鈥淭hat, to me, makes the answer to the question 鈥楬ave they braided non-Abelian anyons?鈥� a yes.鈥� Still, there is the nagging issue of a real-world system to compare against. Yao admits that if we had already made nonabelions in lots of real materials, 鈥渕y gut feeling would be different.鈥�

Now let鈥檚 consider the wormhole incident. What Spiropulu and her colleagues did relates to a long-standing puzzle at the heart of physics: while space and time are described by general relativity as being smooth and continuous, quantum mechanics paints nature as fundamentally discrete and grainy. Can the two views be reconciled in a theory of quantum gravity?

The Ads/CFT correspondence

One of the approaches considered most promising posits that we live in a hologram, based on something called the AdS/CFT correspondence. First proposed in 1997 by theoretical physicist Juan Maldacena, this asserts a relationship between a kind of space-time geometry called an anti-de Sitter space that arises in general relativity and a class of quantum field theories. In effect, it means that if you project the information in the anti-de Sitter space onto its own boundary, that information takes a form that looks a lot like a quantum field theory. The AdS/CFT conjecture is, in other words, a possible way to bridge the chasm between smooth relativistic space-time and the grainy quantum world. One possible implication of this idea is that space-time itself can be considered to be woven out of the web of entangled quantum particles.

Heady stuff. But if the conjecture is right, it is possible to translate the description of a wormhole in space-time as supplied by general relativity 鈥� a shortcut between different regions of space 鈥� into a set of correlations between entangled qubits. Set up such correlations between the qubits of a quantum computer, and AdS/CFT says that this is formally equivalent to a kind of wormhole. You could even transfer 鈥� or 鈥渢eleport鈥� 鈥� information between the entangled qubits in a way that looks equivalent, in this view, to sending it through a wormhole.

This is more or less what Spiropulu and her colleagues did, though they actually used something called the SYK model to set up their qubits, which is a simplified 鈥渢oy鈥� approximation of the AdS/CFT correspondence. They set up their experiment using nine of the qubits on Google鈥檚 Sycamore chip and found that quantum teleportation of information between qubits 鈥渢hrough the wormhole鈥� produced the predicted signal of this equivalence. The title of their paper, , sounds like something out of science fiction: 鈥淭raversable wormhole dynamics on a quantum processor鈥�.

There is quite a difference here from the work on nonabelions and time crystals. Not only is the SYK model a simplified version of the AdS/CFT description of quantum gravity, but furthermore the correspondence between general relativity and quantum field theory supposed by AdS/CFT is purely hypothetical, not least because our own universe isn鈥檛 thought to have anti-de Sitter space-time.

So the issue here isn鈥檛 so much whether making some quantum state from qubits is making the thing itself, but whether that thing bears any relation to physical reality. A wormhole is generally understood to be a part of space-time that enables particular operations, like travelling from here to there in a single leap. Numbers output from a quantum chip don鈥檛 let you do any of that.

鈥淚 think no one would be impressed by claims that scientists had 鈥榗reated a wormhole鈥� if it turned out that all they had done was to program a simulation of a wormhole on a classical computer,鈥� says Aaronson. 鈥淎nd I reject the idea that simulating a wormhole on a quantum computer makes the situation fundamentally different.鈥� New 杏吧原创 invited Spiropulu to reply to these critiques, but she didn鈥檛 respond.

Is reality a simulation?

Wormholes aside, if there is something in the idea of processes in a quantum computer corresponding to 鈥渞eality鈥�, how far could it go? It is a question that has motivated science-fiction writers for years. In the latest series of Netflix show Black Mirror, for example, the 鈥淛oan is Awful鈥� episode features an alternative reality created on a quantum computer that feels totally real to the characters within it.

On the principle that we are physical stuff made of atoms that a quantum simulation would need to be mapped onto, such a computed 鈥渁lternative reality鈥� would indeed be just a simulation. But some researchers have argued that all of reality is, in the end, basically a giant quantum wave function describing patterns of quantum information. From this cosmic perspective, things get a little blurry: what would make one wave function 鈥渕ore real鈥� than another? Indeed, some people believe there is a more than a fair chance this is all we are 鈥� a quantum simulation created by an alien superintelligence.

But then, as Lloyd puts it: 鈥淚f the simulation is indistinguishable from reality, the question of whether we live in a simulation or are just ourselves is not empirically decidable, and we need other criteria to form a judgment.鈥� He suggests that we employ Occam鈥檚 razor, which states that the simplest explanation is usually the best. 鈥淚n this case, Occam鈥檚 razor suggests that we just are ourselves,鈥� he says. Lab-made quantum wormholes may not be real, but there is no good reason to doubt that you are.